EcoSta 2019: Start Registration
View Submission - EcoSta2019
A0353
Title: Development of microbiome-based prediction models using co-informative information Authors:  Michael Wu - Fred Hutchinson Cancer Research Center (United States) [presenting]
Abstract: The low practical and financial cost of collecting microbiome data has spurred interest in the development of microbiome-based models for predicting health outcomes. Many of these studies also collect additional types of genomic data that are co-informative with the microbiome. However, such data are expensive such that the objective is to use these data to facilitate development of a prediction model that only requires collecting microbiome data in the future. To do this, we propose a framework for summarizing the co-informative information within a kernel which can be used to modify the usual LDA loss function. We further consider the imposition of additional practical and biologically informed penalties. Simulations and data applications are used to illustrate the approach.